1. Field of the Invention
This invention relates generally to single in-line memory modules (SIMMs), dual in-line memory modules (DIMMs), and the like, and, more specifically, to SIMM, DIMM and other memory module boards providing at least one extra die site for attachment of an additional die to replace a defective die or dice thereon.
2. State of the Art
An integrated circuit (IC) typically includes a semiconductor die (die) electrically attached to a lead frame providing physical support for the die and connecting the die to external circuitry, such as a printed circuit board or other conductor-carrying substrate. In such an arrangement, the lead frame and die may be connected by wire bonding the lead fingers of the lead frame to contact or bond pads located on a surface of the die. The die and lead frame are then typically encapsulated within a transfer-molded plastic package, although ceramic and metal packages may also be used, depending on the operating environment and the packaging requirements of the die.
As the demand for memory, in particular random access memory (RAM), surpassed the memory capability of a single die, multi-chip modules (MCMs) were developed, such modules having a number of memory devices attached to a single substrate, such as a printed circuit board. A SIMM is a memory module having multiples of the same basic die, where the semiconductor memory chips are aligned in a row and interconnected to a printed circuit board to, in effect, create a single device with the memory capacity of the combined memory chips. An example of a SIMM, including a plurality of dynamic random access memory devices (DRAMs) used as memory in a computer, is illustrated in U.S. Pat. No. 4,992,850, issued Feb. 12, 1991, to Corbett et al., assigned to the assignee of the present invention. As the demand for additional memory on a single device has increased, other devices, such as dual in-line memory modules (DIMMs), have also been developed. Such devices, while providing the desired memory capability on a single printed circuit board, present unique problems for the manufacturer when one or more of the semiconductor memory chips thereon fail.
It is well known that semiconductor dice have an early failure rate, often referred to in reliability terms as xe2x80x9cinfant mortality.xe2x80x9d Moreover, infant mortality of MCMs is multiplied depending on the number of individual semiconductor dice provided therein. For example, a SIMM composed of ten dice, each die having an individual reliability yield of 95%, would result in a first pass test yield of less than 60%, while a SIMM composed of twenty dice, each die having an individual reliability yield of 95%, would produce a first pass test yield of less than 36%.
When a single packaged die, such as a dual in-line package (DIP), fails, a manufacturer can attempt to repair the device, use the device for some reduced capacity function if the device is only partially defective, or scrap it. When complete failure of a die has not occurred and a portion of the memory is good (e.g., 1, 2, or 3 megabits of a 4 megabit chip), such a device is not typically useful. For MCMs such as a SIMM, where a number of semiconductor dice are attached to a single substrate, however, it may not be possible to use the device for some reduced capacity function and it is surely not desirable to scrap the entire MCM when some, if not most, of the dice attached thereto are not defective. Thus, the manufacturer is left with the somewhat costly process of reworking the MCM, typically by removing the defective chips and replacing them with new ones. Such a procedure is described in U.S. Pat. Nos. 5,239,747 and 5,461,544, where a SIMM having a specialized trace pattern suitable for both bum-in and individual die testing is tested to determine if any of the semiconductor devices mounted thereon are non-functional and, if so, the defective device is replaced with a device which has been subjected to burn-in, or the entire multi-chip module can be subjected to another burn-in process after the replacement of the defective device. The defective devices, however, are merely replaced by removing the defective device and replacing it with another, either a device previously subjected to burn-in or not. This rework process can be complicated, time consuming and costly, depending upon the type of device, the type of mounting of the device on the substrate, and the type of substrate used for mounting. For example, plastic-packaged devices are typically physically pulled to disconnect their leads from the module, while so-called xe2x80x9cglob toppedxe2x80x9d (silicone or epoxy gel covering) dice may be removed after cutting through the encapsulant to the wire-bonded die, which is pulled. In addition, since replacing multiple unacceptable dice on an MCM poses physical risks to other MCM dice during the replacement operation, it may be desirable to discard such an MCM rather than attempt rework, particularly where the reliability of the replacement die is not known.
Depending on the extent of testing and/or burn-in procedures employed, a die may typically be classified into varying levels of reliability and quality. For example, a die may meet only minimal quality standards by undergoing standard probe testing or ground testing while still in wafer form, while individual separated or xe2x80x9csingulatedxe2x80x9d dice may be subjected to intelligent burn-in at full-range temperatures with full testing of the die""s circuitry. A die that has been so tested is termed a xe2x80x9cknown good diexe2x80x9d (KGD). Examples of methods for the testing and burn-in of an individual die prior to packaging are disclosed in U.S. Pat. Nos. 5,448,165 and 5,475,317.
A cost-effective method for producing known reliable SIMMs, DIMMs and the like with larger numbers of chips on a single device is desirable for industry acceptance and use. In an attempt to provide known reliable SIMMs complying with consumer requirements, it would be desirable to fabricate the SIMM completely of KGD. Using only KGD in a SIMM, however, would not currently be cost effective since each KGD has to be subjected to performance and burn-in testing, both of which are costly at this point in time. Typically, however, SIMMs are fabricated from probe-tested dice, and are subsequently burned-in and performance tested. In contrast to the use of all KGD in a SIMM, when using dice with well known production and reliability histories, particularly where the dice being used are known to have a low infant mortality rate, the use of such minimally tested dice to produce a SIMM is usually found to be the most cost effective alternative.
As previously stated, since typical testing and burn-in procedures are generally labor and time intensive, posing significant risks to the dice of a SIMM, in the event that a SIMM contains an unacceptable die, replacement of the unacceptable die with a KGD is preferable. Module rework with a KGD does not typically require the SIMM to be subjected to additional burn-in procedures that can unnecessarily stress the dice. An example of a method and apparatus for the testing and burn-in of an individual die prior to packaging is illustrated in U.S. Pat. No. 5,424,652, issued Jun. 13, 1995, to Hembree et al., assigned to the assignee of the present invention. Such a method and apparatus provide a source of KGD to allow for the rework of an unacceptable die in an MCM with a KGD. In other instances, it is known to test a die in a package for functionality and replace any defective die. Such is illustrated in U.S. Pat. Nos. 5,137,836, 5,378,981, and 5,468,655.
One way in the art to eliminate the need to physically remove defective or unacceptable dice from a SIMM has been to provide additional, redundant spaces on the printed circuit board for attachment of replacement chips. Thus, one additional space has been provided adjacent each memory chip on the board, the additional spaces providing contacts for attachment of a semiconductor chip similar to the one it is replacing. For example, if a 32 megabit SIMM contains eight 4 megabit chips, then eight additional spaces are provided on the SIMM, configured to accept up to eight additional 4 megabit chips, if necessary. Such a configuration, however, results in a memory module that is approximately twice as big as a memory module having no extra spaces.
Therefore, a need exists for the cost-efficient fabrication of SIMMs, DIMMs, and the like, of known performance and reliability requirements that requires a minimal amount of rework when one or more dice attached thereto are found defective.
Accordingly, a memory module, such as a single in-line memory module (SIMM) or dual in-line memory module (DIMM), is provided having at least one redundant or auxiliary chip attach location for attachment of a replacement chip. When one or more dice on a memory module are found defective, one or more replacement chips can be attached to the one or more auxiliary chip attach locations with the size of the replacement chips being at least equal to the amount of defective memory. Thus, the defective dice can be replaced without needing to be physically removed. Moreover, by providing auxiliary chip attach locations that can accept different sizes and memory capacities of replacement semiconductor chips, one replacement chip can replace several defective chips on the memory module.
In a preferred embodiment, a SIMM or DIMM board is provided having a plurality of primary chip attach locations and one auxiliary chip attach location. Each of the plurality of primary chip attach locations is similarly configured to accept the same type of semiconductor chip, such as a number of 4 megabit chips. The auxiliary chip attach location, on the other hand, is configured to accept more than one capacity of replacement semiconductor chip. Thus, depending on the amount of defective memory detected on the SIMM, a replacement chip having at least that amount of good memory can be attached to the auxiliary chip attach location. Consequently, the replacement chip may be a 1 megabit chip, if only one 1 megabit of memory is found defective, or a 4 megabit chip, if an entire 4 megabit chip is found to be defective, the configuration of the auxiliary chip attach location being capable of accepting either replacement chip.
In another preferred embodiment, a SIMM or DIMM is provided having two rows of the same type of semiconductor memory chip and a redundant or additional chip attach location for accepting a variety of semiconductor memory chips. The redundant chip attach location is electronically connected in parallel to the rest of the memory chips so that if one or more memory chips are found defective, an auxiliary replacement chip having an amount of memory approximately equal to that found defective can be attached to the additional chip attach location.
Having the capability to easily and cost effectively rework memory modules without the need to remove defective chips or the need to substitute defective chips on a one to one basis is highly desirable. The ability to provide auxiliary chip attach locations that can accommodate a plurality of different chip configurations not only makes rework more simple, but allows memory modules with large numbers of chips to be cost effective. For example, in yet another preferred embodiment, a memory module having three rows of similar memory chips is provided with more than one auxiliary chip attach location, each of the auxiliary chip attach locations being capable of receiving more than one type of semiconductor die. Thus, if only one of the many chips provided thereon fails entirely, then one substantially identically configured chip can be attached to any one of the three auxiliary chip attach locations. If more than one memory chip fails or more defective memory is located than can be replaced with a single auxiliary chip then, if necessary, more than one replacement chip, can be attached to any one or more of the three auxiliary chip attach locations.
In yet another preferred embodiment, rather than having all of the auxiliary chip attach locations capable of receiving variously configured chips, at least two auxiliary chip attach locations are each provided with different configurations. Thus, depending on the number of bad memory chips, one or more chips having a combined memory capacity substantially equal to the bad memory can be attached to the auxiliary chip attach locations. For example, on a DIMM with five 4 megabit chips found defective, a replacement 16 megabit chip can be attached to the auxiliary chip attach location configured to receive 16 megabit chips and a replacement 4 megabit chip can be attached to the auxiliary chip attach location configured to receive a 4 megabit chip.
Preferably, the replacement chips are KGD so that the additional burn-in is not required on the memory module. Moreover, the KGD may be partially defective dice or xe2x80x9cpartialsxe2x80x9d that are known to be good for a certain capacity of memory (e.g., 3 megabits of a 4 megabit chip). This is particularly attractive, since a high percentage (approaching 50%) for some designs of 16 megabit DRAMS is partially or completely defective, while 5-10% of 4 megabit DRAMS comprise partials. Since partials might otherwise be discarded, beneficial use thereof as replacement chips enhances the effective yield rate for the chips and lowers per-unit-memory costs.
Thus, for example, for SIMMs having a design memory capacity of 32 megabits with 7 of the 32 megabits tested defective, a replacement partial 16 megabit chip known to be good for 7 megabits could be attached to an auxiliary location. The defective memory of 7 megabits on the primary chip(s) could then be disabled. Similarly, for a 32 megabit memory module with 7 bad megabits of memory and more than one auxiliary chip attach location, a replacement full 4 megabit KGD and a 3 megabit xe2x80x9cpartialxe2x80x9d KGD could be attached to two of the auxiliary chip attach locations.